The present invention is directed to, inter alia, passive correction of turbulence affected incoherent imaging using an optical system (and methodology) for, inter alia, reducing the effects of short-exposure blur due to atmospheric optical turbulence.
Optical signals passing through a time varying inhomogeneous medium, such as the Earth's lower atmosphere, can become significantly distorted when propagating over ranges of even as short as several hundred meters. The primary mechanism of this distortion is due to temperature fluctuations driven by heating and cooling of the air which is most severe at the Earth's surface. In such cases, several optical distortion effects impact propagating optical waves and signals. Coherent propagation is significantly affected by turbulent scintillation (amplitude fluctuation) effects, and beam wander of propagating laser beams. For incoherent wave sources being viewed by passive imaging systems, three effects occur: Short exposure images show blurring of point sources in the object plane. Point objects also appear to wander in position due to angle-of-arrival variations. Thirdly, point sources separated by angular distances exceeding a characteristic value (the isoplanatic angle) appear to wander independently.
Systems for correcting for turbulence effects on imaging through use of an imaging system containing a Spatial Light Modulator (SLM) to modify the phase of the incoming radiation have been disclosed, such as, for example, M. A. Vorontsov, G. W. Carhart, and J. C. Ricklin, “Adaptive phase-distortion correction based on parallel gradient descent optimization,” Opt. Lett. 22, 907-909 (1997) (hereby incorporated by reference) and G. W. Carhart, J. C. Ricklin, V. P. Sivokon, and M. A. Vorontsov, “Parallel perturbation gradient descent algorithm for adaptive wavefront correction,” in Adaptive Optics and Applications, R. Tyson and R. Fugate, eds., Proc. SPIE 3126, 221-227 (1997) (hereby incorporated by reference). These methods described in these papers focus on two aspects: the use of a Spatial Light Modulator (or deformable mirror, which functions in a similar fashion) and use of the Parallel Gradient Descent (PGD) optimization method, which has now evolved to be known as the Stochastic Parallel Gradient Descent (SPGD) method.
Ground-to-ground imaging-through-turbulence problems essentially involve image blur, which is caused primarily by turbulence close to the system receiving aperture, and image distortion, which is due to turbulence weighted toward the target object that is under observation by the system. Because the atmospheric optical turbulence strength is greatest close to the ground, the dominant effect impacting ground-based imaging systems is turbulence-induced blur close to the system aperture. As used herein, the terminology “blur” signifies to make indistinct and hazy in outline or appearance, reducing high angular-frequency detail, as opposed to obscuration which affects contrast, which reduces all angular-frequency detail equally. As used herein, the terminology “distort” as it relates to optics means a change in the shape of an image resulting from variations in the perceived relative angular positions of different scene features of a given viewed object.
To correct for image blur, various system configurations have been proposed in past work. These usually involve active system implementations that feature some sort of illumination device to produce what is commonly known as a “guide star.” A guide star is a compact illumination source of known geometry and phase that can be imaged through the turbulent atmosphere by the imaging system. The system then analyzes the propagated characteristics of this guide star and uses the results of this diagnosis to formulate a correction to the optical system. This correction always involves a deformable optical device, either a deformable mirror or a spatial light modulator (SLM). The guide star can be formed by an illumination beam propagated by the system itself, producing an illuminated spot in the object field of view, or by an illumination source placed in the imaged object field of view and oriented toward the receiver optics. The following patent materials rely on the use of guide stars and/or the use of a wave front sensor: U.S. Published Application No. 2004/0208595, Free Space Communication System with Common Optics and Fast, Adaptive Tracking, by Fai Mok and James Kent Wallace; U.S. Published Application No 2005/0045801, State Space Wavefront Reconstructor for an Adaptive Optics Control, by Carey A. Smith; U.S. Published Application No 2006/0049331, Adaptive Optics Control System, by Carey A. Smith; U.S. Pat. No. 7,038,791, Signal-to-Noise Ratio Tuned Adaptive Optics Control System, by Carey A. Smith; U.S. Published Application No 2010/0080453 A1, System for Recovery of Degraded Images, by Nicholas George; and U.S. Pat. No. 6,163,381, Dual Sensor Atmospheric Correction System, by Donald W. Davies, Mark Slater, and Richard A. Hutchin.
Unfortunately, there are several problems with the use of guide stars. First, a guide star approach is not a passive solution. Active systems that require the illumination of a target scene prior to detection of significant targets are not stealthy and are undesirable in most tactical situations that are of interest in a military situation. Second, many imaged objects may not have useful reflective properties that will work properly with an illumination beacon. To provide a proper guide star an object would need to have a corner reflecting or shallow convex specular surface (a “glint” target) nearby. Most natural objects are diffuse reflectors and thus do not return glints. Many man-made objects are also diffuse reflectors or have specular surfaces that are sharply curved and thus only return a very weak glint. Other alternatives, such as placing an illuminator in the object plane, requiring objects of interest to mount glint reflectors, or forming laser-induced fluorescence (LIF) guide stars on target surfaces are obviously not practical from an Army application standpoint. Another difficulty with the guide star approach is that the coherent propagating wave from a guide star is affected by turbulent scintillation, which is most strongly weighted at the center of the optical path, not at the system receiver. Thus the guide star method is not optimized to produce a useful result for removing turbulent blur.
Therefore, a means is needed to image objects through turbulent blur that does not require an active illumination beacon (a guide star) and is optimized to sense turbulent blur perturbations on imaged incoherent radiation.
As opposed to active wave front sensing techniques, U.S. Published Application No 2005/0151961, Surface Layer Atmospheric Turbulence Differential Image Motion Measurement, by John T. McGraw, Peter C. Zimmer, and Mark R. Ackermann simply attempts to sense the image distorting effects of the atmosphere without actually attempting to modify or correct for turbulence effects.
Two known exceptions to the general approach of active wave front sensing are Patent 2010/0053411 A1, Control of Adaptive Optics based on Post-Processing Metrics, by M. Dirk Robinson and David G. Stork, and the method proposed in G. W. Carhart, J. C. Ricklin, V. P. Sivokon, and M. A. Vorontsov, “Parallel perturbation gradient descent algorithm for adaptive wavefront correction,” in Adaptive Optics and Applications, R. Tyson and R. Fugate, eds., Proc. SPIE 3126, 221-227 (1997). Both of these propose a system or a method to performs wavefront correction based on post-processing of received imagery to produce a metric that is then used in guiding the correction of images. The former proposed a system. The latter proposed a processing approach based on an algorithm. Both based their corrections on image analysis alone.
As indicated above, the guide star concept is generally not preferred in ground-to-ground imaging applications for blur correction. In assessing the impact of turbulence on boundary layer imaging, two observations are manifest. First, the turbulence that is causing the most image blur is close to the sensing aperture. Second, scene elements that are separated in the scene by a significant angular separation experience anisoplanatic effects limiting the ability of a system to correct turbulent image perturbations at large angular separation from the guide star. Anisoplanatism means that turbulent perturbations in different parts of the atmospheric field are causing different turbulent perturbations in different parts of an imaged scene. This effect impairs the performance of guide-star-based systems, because turbulent perturbations on the guide star wavefront in one part of the image frame are not the same turbulent perturbations that impact scene elements in another part of the image frame. Guide-star-based systems thus do not do well at correcting for turbulent blur in different parts of an imaged scene, underscoring the need for a passive method that can correct for turbulence sequentially in different parts of the image.
Unlike systems that rely on a coherent “guide star” signal to provide a sufficient density of photons to feed a wavefront detection process, atmospheric boundary layer imagers typically observe light emerging from a plurality of decorrelated emission sources. In particular, source points present on surfaces that are rough on the order of a single wavelength of the propagating radiation, will not produce a single coherent source even in a point source sense. In this instance a second type of solution to compensating for turbulence has been sought. This second form of solution involves a purely passive approach, of varying implementations, generally involving one or more post-imaging processing procedures for detected signals to remove the impacts of turbulence. One such algorithmic approach entails dewarping of the imaged field, to remove image distortion, based on analysis of the temporally varying apparent positions of objects. Another approach uses long-term averaging of images to essentially remove the angle-of-arrival variations followed by inverse filtering using a best-guess of the long-exposure atmospheric MTF. Various combinations of these two approaches can be constructed, including analysis of and segmentation of images to separately study constant regions of images and regions that are considered to be temporally evolving that may contain scene elements of interest. Many of these procedures fall under the category termed the “lucky pixel” or “lucky patch” method, initially proposed by D. L. Fried, “Probability of getting a lucky short-exposure image through turbulence,” J. Opt. Soc. Am., 68:1651-1658 (1978). However, whereas Fried's initial proposal suggested capturing complete distortion free images, later implementations of this concept [e.g. Carhart, G. and M. Vorontsov, Opt. Lett. 23, 745-747 (1998) or Vorontsov, M., JOSA A 16, 1623-1637 (1999)] first segment the images into a series of sections, analyze each section separately to determine the relative clarity (spatial frequency content based on an image quality metric) of each, and then, on a section-by-section basis, proceed to composite a complete reduced-turbulence equivalent image as a mosaic. Unfortunately, for many terrestrial (ground-to-ground) imaging scenarios the probability of obtaining any portion of an image that is free of significant turbulence may be so small as to provide a negligible chance of obtaining a set of null-turbulence patches sufficient to construct an unperturbed image. One means of rapidly evaluating the overall quality of either an image portion or a complete imaged scene involves constructing a sum of squares of the normalized image scene pixels [N. Mahajan, J. Govignon, and R. J. Morgan, “Adaptive optics without wavefront sensors,” SPIE Proc., 228, Active Optical Devices and Applications, 63-69 (1980)]. As Mahajan et al. explained, the variance of the image information is related to the area under the combined atmosphere plus system MTF. Using this metric, it is possible to gauge the level of spatial frequency energy in a scene. Also, while D. L. Fried, “Probability of getting a lucky short-exposure image through turbulence,” J. Opt. Soc. Am., 68:1651-1658 (1978) focused on only the probability of detecting a lucky short-exposure image of a scene, a later study by R. E. Huffiagel, “The Probability of a Lucky Exposure,” Tech. Memo. REH-0155, The Perkin-Elmer Corp. (1989) (hereby incorporated by reference) considered how this capability is enhanced for signals that have been partially corrected through the removal of an increasing number of phase perturbation aberration modes. In particular, Hufnagel considered cases involving 0 (short-exposure image only), 3 (2nd order aberrations), 7 (2nd and 3rd order aberrations) and 12 (2nd through 4th order aberrations corrected) modes compensated. Hufnagel's study, interpreted for the lucky patch problem, indicates that while the lucky patch method acting on a short-exposure image series could provide significant improvement for cases where the ratio of the diameter of the optics to the coherence radius (X=D/r0) is no more than three, by correcting 7 aberration modes, the lucky patch correction technique would be one million times more likely to find a lucky patch at X=10. This implies that the lucky patch approach could be extended to turbulence conditions ten times stronger (triple the range) of the baseline lucky patch method alone.
In U.S. Published Patent Application No. 2010/0053411, entitled “Control of Adaptive Optics based on Post Processing Metrics,” by M. Dirk Robinson and David G. Stork, a system is proposed that performs wavefront correction, based on post-processing of received imagery to produce a metric that is then used in guiding the correction of images. The system appears to only apply to static targets.
However, it would appear that if one attempts to correct imagery based solely on analysis of received imagery, the number of phase perturbation aberration modes (expressed in terms of Zernike expansion functions [e.g. V. N. Mahajan, “Zernike annular polynomials and optical aberrations of systems with annular pupils,” Appl. Opt. 33:8125-8127 (1994) or G.-M. Dai and V. N. Mahajan, “Zernike annular polynomials and atmospheric turbulence,” J. Opt. Soc. Am. A 24:139-155 (2007)] present at the system aperture gives rise to a problem. That problem is a limitation on how frequently a given mode may be corrected given a specific rate of image collection by the optical system, in combination with the strength of aberration due to a specific mode. To assess the effective number of active perturbation modes present one must be able to evaluate the statistical state of the wavefront present at the system aperture. D. L. Fried, “Optical resolution through a randomly inhomogeneous medium for very long and very short exposures,” J. Opt. Soc. Am. 56:1372-1379 (1966) suggested that an appropriate measure of the decorrelation in the phase front present in the system aperture is the turbulent coherence diameter, designated r0. This length is a width measured in a plane transverse to the direction of wave propagation over which the wave phase coherence decays by a value of exp(−1). For many common long-range surveillance receiving systems imaging objects at several kilometers distance the wavefront will become decorrelated within the diameter of the receiver aperture at even moderate optical turbulence levels (characterized by the dimensionless ratio X=D/r0, where D is the diameter of the system aperture). In the case where X significantly exceeds unity, the wave will exhibit random phase fluctuations that can cause image blurring effects, even accounting for short-exposure imaging [Tofsted, D. H, “Reanalysis of turbulence effects on short-exposure passive imaging,” Opt. Eng., 50:01 6001 (2011) (hereby incorporated by reference)]. To describe these decorrelations in the propagating phase front, various orthonormal families of basis functions may be utilized [e.g. V. N Mahajan, “Zernike annular polynomials and optical aberrations of systems with annular pupils,” Appl. Opt. 33:8125-8127 (1994). Fried, “Statistics of a Geometric Representation of Wavefront Distortion,” J. Opt. Soc. Am. 55:1427-1431 (1965)]. As the ratio X exceeds unity, the effective number of non-zero expansion modes needed to describe the wave perturbation function increases approximately as X2. Therefore, for any given degree of turbulent perturbation one must have a specific plan to enable phase corrections [see detailed discussions of FIGS. 5, 11, and 14 through 18] based on an organized methodology.
A further example of prior art is the application of the stochastic parallel gradient descent technique in tracking the effects of turbulent fluctuations. This stochastic method attempts to adjust a sequence of deformable mirror pistons by performing random fluctuations of the current choice of piston settings, and adaptively modifying the best guess of the correction state based on the outcome of each stochastic perturbation. Weighting methods may be used to selectively focus the algorithm on the correction of lower order modes. The limitation of this approach is the high number of image samples to be collected rapidly enough (several thousand sample images per second) to track the evolving state of the various perturbation modes. This is because the method is relatively inefficient, relying on a stochastic adjustment procedure. Because the maximum sampling rate of an image at an adequate signal-to-noise ratio is limited by the amount of ambient light available to produce the image and the system's light gathering capability, sufficiently high frame rates may not be possible without the augmentation of the system by a high intensity light source in the imaged scene to provide the necessary illumination. This amenity may not be available in military or in many other contexts.